pymc-devs / pymc-devs/pytensor-ml
Test groups have drifted out of balance, so CI wall clock is 8 minutes
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 9
- Forks
- 7
- Avg merge
- 6h 55m
- Merged PRs (30d)
- 40
Description
The ubuntu groups were balanced on cold-compile time when they were set up, but layers and optim have grown into it since — layers alone now takes 4x what attention and rest do, and it sets the wall clock by itself. The windows smoke job is the other pole, 7m29 for three files.
layers hasn't been rebalanced since it was defined in #64, and five test files landed in it between #92 and #103, so that's probably most of the drift.
layers 8m04 ubuntu
smoke 7m29 windows
optim 4m56 ubuntu
dispatch-jax 2m53 ubuntu
dispatch-torch 2m27 ubuntu
dispatch-mlx 2m15 macos
rest 1m51 ubuntu
attention 1m50 ubuntu
(latest completed run on main; --durations=25 is already on, so the per-test breakdown is in each job's log)
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by inspecting the CI job definitions for the Ubuntu test groups and the Windows smoke job, then review the per-test breakdown from the latest completed run. Reassign the slowest tests so the groups have more balanced durations, including the five files added between #92 and #103. Done means the group wall clocks are materially closer and the test suite still passes on the affected platforms.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- ci-cd, performance, testing-qa
- Issue type
- Refactor
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 68/100